The latest daily build has the following smatch issues: drivers/media/i2c/ccs/ccs-data.c:524 ccs_data_parse_rules() warn: address of NULL pointer 'rules' Sakari has a patch for that, but it fails some tests and it needs more investigation. drivers/staging/media/atomisp/i2c/atomisp-ov2680.c:476 ov2680_init_cfg() error: we previously assumed 'sd_state' could be null (see line 468) Hans de Goede is working on dropping atomisp-ov2680.c and using the mainline ov2680 driver instead, so once that's in this issue will disappear. There is the following sparse warning: drivers/media/usb/siano/smsusb.c:53:38: warning: array of flexible structures The siano driver embeds a struct urb in a larger struct, and that's wrong. It should be allocated instead. Is there anyone who can look at this? Mauro, do you still have siano hardware to test with? There is also one warning indicating that two kernel configs are never compiled: Check COMPILE_TEST: WARNINGS: VIDEOBUF_VMALLOC VIDEOBUF_DMA_CONTIG This is pending the bttv driver vb2 rework: once that is merged, the old videobuf can be removed and this warning will disappear. It's waiting for me to do final (?) testing of the v3 of the bttv series. Note that both sparse and smatch are updated to their latest versions whenever a daily build starts, so if new checks are added to either project, then a new set of warnings/errors can appear. It is, however, important to try and keep this as clean as possible since while some are false positives, in most cases they catch actual bugs. If too many issues are reported, then it becomes hard to detect when a new patch introduces new issues, and bugs will slip through (this happened to me just a few weeks back). If working around a false positive requires too much code, then is it best to reach out to the sparse/smatch developers to see what can be done to improve the check. Regards, Hans
